Genetic disruptions of Drosophila Pavlovian learning leave extinction learning intact

Abstract

Individuals that experience traumatic events may develop persistent post-traumatic stress disorder (PTSD). Patients with this disorder are commonly treated with exposure therapy, which has had limited long-term success. In experimental neurobiology, fear extinction is a model for exposure therapy. In this behavioral paradigm, animals are repeatedly exposed in a safe environment to the fearful stimulus, which leads to greatly reduced fear. Studying animal models of extinction already has lead to better therapeutic strategies and development of new candidate drugs. Lack of a powerful genetic model of extinction, however, has limited progress in identifying underlying molecular and genetic factors. In this study, we established a robust behavioral paradigm to study the short term effect (acquisition) of extinction in Drosophila melanogaster. We focused on the extinction of olfactory aversive one-day memory with a task that has been the main workhorse for genetics of memory in flies. Using this paradigm, we demonstrate that extinction can inhibit each of two genetically distinct forms of consolidated memory. We then used a series of single-gene mutants with known impact on associative learning, to examine effects on extinction. We find that extinction is intact in each of these mutants, suggesting that extinction learning relies on different molecular mechanisms than does Pavlovian learning.

Introduction

Pavlovian fear conditioning and it’s extinction in animals have been widely used as behavioral models to uncover mechanisms of memory generally and of human anxiety disorders in particular. In the Pavlovian paradigm, animals are simultaneously exposed to a conditioned stimulus (CS+, e.g. light, tone or odors) and an unconditioned stimulus (US, e.g. electric foot shock). The animals learn to associate the CS+ as a predictor of the US and exhibit a conditioned fear response (CR, e.g. freezing or avoidance behavior) to the CS+. This conditioned fear response can be weakened by a process called extinction in which the conditioned animals are subjected to repeated CS+ exposure without US reinforcement. Extinction, which first was studied by Pavlov (Pavlov, 1927), has since formed the basis for development of behavioral therapy in humans.

The evidence from animal models and from human subjects strongly supports the view that extinction involves an active learning process rather than simply an erasure of the fear memory (Bouton, 2002, Quirk & Mueller, 2008). Two types of findings support this conclusion. First, the learned fear can return following extinction either with the passage of time (spontaneous recovery), with re-exposure to the US (reinstatement) or with exposure to the context in which the fear conditioning was acquired (renewal). The return of the behavioral manifestations of fear creates a challenge for therapy in humans and supports the conclusion that extinction includes an active process that masks but does not erase the conditioned fear. A second line of evidence in support of the above notion derives from genetic and pharmacological manipulations that can disrupt extinction allowing the conditioned fear response to re-emerge.

Like other types of learning, extinction appears to consist of acquisition, consolidation and retrieval. As with consolidation of associative memory, consolidation of extinction can include a requirement for a cascade of new gene expression (Barad & Cain, 2007). This shared requirement is consistent with the view that similar cellular mechanisms might be used to consolidate conditioned memory and extinction memory. On the other hand, pharmacological and genetic studies in animal models have identified molecules unique to extinction as well as molecules common for the consolidation and retrieval of extinction memory and conditioned memory (Myers & Davis, 2007). Indeed findings from animal models have led to promising drugs to facilitate the relief of anxiety disorders (Quirk & Mueller, 2008). As an example, DCS (D-cycloserine), a partial NMDA receptor agonist, was shown in animal studies to facilitate the consolidation of extinction, likely through MAPK and PI3K (Lu et al., 2001, Szapiro et al., 2003, Walker et al., 2002, Yang & Lu, 2005). In later clinic studies, DCS has been reported to facilitate exposure therapy for acrophobia (Ressler et al., 2004) and social anxiety disorder (Hofmann et al., 2006).

Compared with consolidation, the acquisition of extinction (extinction learning) is less well-studied. Although a number of genes have been reported to play roles in acquisition of extinction, the underlying molecular mechanism are far from being understood (Busquet et al., 2008, Cain et al., 2002, Kamprath et al., 2006, Kim et al., 2007, Marsicano et al., 2002, Mckinney et al., 2008, Mcnally et al., 2005, Mcnally et al., 2004, Mcnally & Westbrook, 2003, Schafe, 2008, Sotres-Bayon et al., 2007).

We chose to study extinction learning in a genetic model organism Drosophila melanogaster. A long history of behavior genetic studies in this model organism already has lead to identification of a large number of single-gene mutations that impact Pavlovian aversive memory and learning (Heisenberg, 2003, Keene & Waddell, 2007, Margulies et al., 2005, Mcguire et al., 2005). Much also is known about the neural circuitry and underlying cellular mechanisms that support Pavlovian olfactory learning in Drosophila. This rich history potentially provides a foundation for detailed mechanistic investigation of the circuits and cell biological mechanisms of extinction. To date, however, only three studies have examined extinction of classical olfactory conditioning in Drosophila (Lagasse et al., 2009, Schwaerzel et al., 2002, Tully & Quinn, 1985). Two of these studies focused on extinction at an early time-point after the standard odor-electro-shock conditioning procedure (Schwaerzel et al., 2002, Tully & Quinn, 1985) and the third investigated extinction after consolidation using an odor-mechanical shock procedure (Lagasse et al., 2009).

To make use of the large collection of genetic reagents that have been established to probe mechanisms of Pavlovian conditioning, we used the electric shock-mediated procedure (Tully & Quinn, 1985) in which most of these mutants have been studied. We first established the parameters for an assay of extinction acquisition after allowing consolidation of the Pavlovian memory to take place. This largely avoided the confounding effects of interaction between acquisition of extinction and decay of the CS-US association. We then used this behavioral assay, to measure extinction acquisition with a panel of mutant strains that are defective in a diverse array of cellular signaling pathways that have established roles in olfactory memory association. We report that none of these mutants have detectable impact on acquisition of extinction learning. Together, our results are consistent with the view that extinction learning relies on distinct molecular/cellular mechanisms from those that underlie aversive Pavlovian conditioning.

Materials and Methods

Fly stocks

Flies used in this study included w¹¹¹⁸(isoCJ1) and w¹¹¹⁸(CS7) wild type control strains, rutabaga1, dNR1(EP331), dNR1(EP3511), radish1 and rugose1 mutant alleles. w¹¹¹⁸(isoCJ1) and w¹¹¹⁸(CS7) are isogenic white lines derived from Canton-S. All mutant strains were first equilibrated to w¹¹¹⁸(isoCJ1) for at least 5 generations. In the case of rutabaga and rugose, which are X linked, we equilibrated the autosomes to that of the control strain . Flies were cultured in standard fly food and room temperature (22.5 °C).

Behavior

All behavior experiments were done in an environmental chamber with red light, 25 °C, 70% humidity. For the experiments using more than two genotypes, the experimenter was blind to the genotype during behavioral experiments and all experiments within a panel were conducted in parallel in a genotype and treatment balanced fashion.

Pavlovian Conditioning

For short-term memory, flies were trained with a single training session as described in Blum et al. (2009). For long-term conditioning, flies were first given either 10 repetitive spaced or 10 repetitive massed training sessions as described in Tully et al. (1994). Flies then were transferred into food vials and incubated at 18 °C until testing.

Extinction training

Extinction training was performed in the same apparatus as used for spaced and massed training (Tully et al. 1994), except that here the training chamber did not contain electric grids. Each single extinction training session formed part of a block of training sessions that are identical to a massed Pavlovian training session except for the lack of grid surface or foot shock reinforcement. During each such session, flies were exposed sequentially to the two odors for 60 seconds each with a 45 second air exposure in between. The CS+ exposure was given first, followed by the CS- (3-octanol or 4-methyl-cyclohexanol), 45 seconds rest in between. Each extinction training session then was repeated with 45 seconds interval to form a massed extinction training program. 45 seconds after the last extinction session, the flies were transferred to the standard T-maze (Blum et al., 2009, Dubnau et al., 2003, Tully & Quinn, 1985) to test memory retention. Air exposure control animals were trained and tested in parallel to each extinction training group. The training procedure for the air control group was identical to extinction training except that the flies were exposed to ambient air instead of odors.

Statistics

According to central limit theorem, the PIs in this study (the average of two percentages) should be normally distributed (Sokal & Rohlf, 1995, Tully & Gold, 1993). Raw data were analyzed with JMP5.1 (SAS Institute Inc.). For Figure1A, PIs were subjected to One-Way ANOVA with training cycle as the main factor. For Figure1B, PIs were subjected to a One-Way ANOVA test. The later comparisons (Tukey-Cramer test) were deemed significant if p<0.05. For Figure 2, PIs were subjected to Two-Way ANOVA with extinction training, memory type (massed vs. spaced or 1 day vs. 4 days) as two main effects and Extinction training X Memory type as interaction term. Later planned comparisons were deemed significant if p<0.05. For Figure3, PIs were subjected to Two- Way ANOVA with Extinction Training and Genotype as two main effects and Extinction Training X Genotype as interaction term. Later planned comparisons were deemed significant if p<0.05. For Figure 4, PIs were subjected to One-Way ANOVA with genotype as main effect. Later planned comparisons were deemed significant if p<0.05.

Figure 1.

Extinction of memory 24 hours after spaced training. (A) Different numbers of extinction training cycles have a graded impact on memory performance of WT flies. 5 (5X) cycles of CS-odor exposure did not yield significant extinction (n=7, p>0.07), but 15 (15X) or 30 (30X) cycles of exposure each led to significant extinction, reducing conditioned avoidance by approximately 50% relative to air exposed controls (n=7 for 15 cycles, p<0.001; n=12 for 30 cycles, p<0.0001). (B) The reduced conditioned avoidance observed with 15 or 30 cycles of extinction are not due to habituation or to sensory adaptation. 30 (30X-odor) or 60 (60X-odor) rounds of odor pre-exposure given to naïve flies did not significantly reduce subsequent Pavlovian conditoning (n=7, p>0.05) relative to either air pre-exposed groups (30X-air, 60X-air) or to animals that were not pre-exposed to air or to odors (no exposure). Even with a massive exposure to 90 cycles of odor pre-exposure (90X-odor), only a relatively mild reduction in subsequent learning ability was observed compared to air pre-exposed (90X-air) or animals not given pre-exposure (no exposure) (n=8, p<0.05). In all the figures, the values represent means and error bars are the standard error of the mean.

Figure 2.

Extinction effectively inhibits two different forms of consolidated memory. Conditioned avoidance was inhibited by 30 cycles of extinction given 24 hours after either massed (Massed training) or spaced training (Spaced training) (A) (n=8, p<0.05) as well as 4-days (4 days) after spaced training (B). Thus extinction can inhibit the consolidated memory known as ARM, which forms after massed training. Memory 24 hours after spaced training is thought to be composed of both ARM and LTM. But ARM decays away within 4-days after training (Tully et al., 1994) leaving a pure measure of LTM. Like ARM, LTM also is inhibited by extinction because 30 cycles of odor exposure inhibit performance both 24-hours (Spaced training, A and 1 day, B) and 4-days (4 days, B) after training (n=23, p<0.05). Interestingly, the magnitude of the extinction effect was significantly greater at the 1-day (1 day) time-point than at the 4 day (4 days) retention time point (F(1, 91)=8.5, p<0.05). Error bars are the standard error of the mean.

Figure 3.

Mutations that disrupt Pavlovian learning leave extinction acquisition intact. rutabaga1 (A, B) and rugose1 (C), dNR1(EP3511) (D, E), dNR1(EP331) (F), radish1 (G, H) mutants were subjected to extinction training one day after spaced Pavlovian conditioning. In the case of rutabaga1 (A, B), dNR1 (D, E and F) and radish1 (G, H), we tested both 30 and 10 cycles of extinction. For rugose1 we tested effects of 10 cycles of extinction (C). With each mutant, we observe significant inhibition of conditioned avoidance after extinction training (n>8; p<0.05). A TWO-WAY-ANOVA analysis shows no interactions between genotype and extinction training (n>8; p>0.05). Error bars are the standard error of the mean.

Figure 4.

rutabaga and dNR1 pathways together account for most of the performance after Pavlovian olfactory learning. The rutabaga1 (rut1/Y) and dNR1 mutants (rut1/+; dNR1(EP331)/dNR1(EP3511)) each cause a partial but significant reduction in learning performance measured 3-minutes after one Pavlovian training session. Performance of the double mutant (rut1/Y; dNR1(EP331)/dNR1(EP3511)) flies was significantly lower than that of each single mutant (rut1/Y or rut1/+; dNR1(EP331)/dNR1(EP3511)) (n=18, p<0.05). Because the rutabaga1 allele is a functional null, these data suggest that dNR1 mainly impacts a rutabaga independent pathway for Pavlovian memory. Error bars are the standard error of the mean.

Results

A behavior paradigm to study extinction of consolidated aversive olfactory memory

Drosophila melanogaster can be conditioned to associate foot shock (US) with odor (CS+) by using a Pavlovian conditioning paradigm (Tully & Quinn, 1985). Genetic and behavioral studies with this assay have identified a series of mechanistically distinct phases of memory consolidation. Short-term and middle-term memory (STM and MTM) form rapidly and require only one training session, but these early memories are unstable and decay rapidly. Memory after one training session is gone within one day. In contrast, repeated training can induce longer-lived forms of memory that can be stable for several days. Consolidated memory in this model system has been further dissected into at least two distinct forms (Tully et al., 1994, Yin et al., 1994), for review see (Margulies et al., 2005). Massed training, which involves repeated CS-US pairings with no rest interval between trials, induces a stable memory often called anesthesia resistant memory (ARM). ARM can last for several days, and is resistant to pharmacological and genetic disruptions of new gene expression. In contrast, memory after spaced training induces an additional consolidated form of memory, which we refer to as LTM (Tully et al., 1994) but cf (Isabel et al., 2004). Spaced training-induced consolidated memory requires induction of a CREB-mediated cascade of gene expression (Yin et al., 1994).

Early attempts at extinction in Drosophila focused on an early time-point when the memory of the original aversive conditioning still is unstable (Schwaerzel et al., 2002, Tully & Quinn, 1985). This creates a confound both because the original memory still is labile to disruption and because memory at these early stages continues to decay rapidly during the time needed for extinction training. In this study, as in (Lagasse et al., 2009), we instead focused on the extinction of one day memory that is already consolidated. At the 24-hour time-point, the conditioned aversion is stable to disruption and to the passage of time. This allows the effects of extinction and the decay of the conditioned aversion to be dis-entangled. Because the panel of genetic reagents has been characterized using the shock-mediated training, we adapted the extinction training to this Pavlovian paradigm.

We first used wild type flies to establish a behavior paradigm to study the acquisition of extinction. Wild-type (CS derivative isoCJ1) first were trained to form aversive memory using our standard spaced training procedure (Blum et al., 2009, Dubnau et al., 2003, Tully et al., 1994). In this spaced training procedure, animals are given 10 training sessions, interspersed with a 15 minute rest interval (see methods). One day after spaced training, we then exposed these flies to varying numbers of extinction trial in which the two odors were sequentially presented (CS+ and CS− alternately) without foot shock reinforcement. Our rationale for presentation of both odors rather than just the CS+ is that the Pavlovian assay also is a discriminative one in which one odor is explicitly paired with reinforcement and the other is explicitly unpaired. In fact, because the avoidance measured during memory retrieval itself involves a discrimination between the two odors, our use of both odors during extinction also minimizes the potential impact on performance of habituation, sensory adaptation or other non-specific effects. After this extinction procedure, we immediately tested memory of the odor shock association.

With this procedure, we observe significant reduction in the avoidance of the CS+ relative to the CS− and this extinction is graded as a function of the number of extinction trials. As shown in Figure 1A, 5 cycles of extinction training did not significantly inhibit the 24-hour memory performance (n=7, p>0.07), whereas 15 or 30 rounds of extinction training resulted in approximately a 50% reduction of the conditioned avoidance (n=7, p<0.001 and n=12, p<0.0001 respectively). We also tested whether the residual conditioned avoidance could be extinguished by increasing the number or extinction trials and found that even after 60 or 90 cycles of extinction training, further reduction in conditioned avoidance was not seen (data not shown). We thus reasoned that the effect of extinction training reaches asymptotically high levels within 30 cycles.

The discriminative design of our procedure should prevent the apparent appearance of extinction because of trivial effects such as reduced sensation of the CS+ relative to the CS−. But we also wanted to rule out the possibility that the reduced avoidance was due to severe sensory adaptation or habituation to both odors because of repeated odor exposure. To exclude this possibility, we conducted a series of control experiments in which we first exposed naive flies to 30, 60 or 90 cycles of odor presentation (CS+ and CS− alternately as in the extinction trials). These pre-exposed animals then were trained in the standard Pavlovian procedure and tested for conditioned avoidance immediately afterwards. 30 cycles of odor pre-exposure did not affect the flies’ ability to form the odor-shock association and avoid the paired odor compared to either naïve flies or flies pre-exposed to air flow (n=7, p>0.05 ) (Figure 1B). Because we observe robust extinction with 10 (see below), 15 and 30 extinction trials (Fig 1A), this essentially rules out impacts of sensory adaptation or habituation on our measured extinction effect. Moreover, even after 90 cycles of odor pre-exposure, flies were still capable of forming short-term memory although their performance was slightly but significantly lower than control groups. The subtle impact of this massive odor pre-exposure, may result either from sensory adaptation, habituation or latent inhibition. Together, these results strongly argue that the reduced performance observed with the 15 and 30 cycles of extinction training are due to inhibitory extinction learning, not sensory-motor fatigue.

Extinction after spaced versus massed training

Genetic and behavioral manipulations in Drosophila have identified two parallel forms of consolidated memory that rely on distinct mechanisms. Repetitive training with no-rest interval (massed training) induces a long-lived form of memory that is anesthesia resistant (ARM) and also resistant to inhibitors of protein synthesis. This memory after massed training, which can last for more than 24 hours, does not appear to require CREB-mediated transcription, but is disrupted in radish mutants. In contrast, repetitive spaced training induces not only ARM, but also a protein synthesis dependent LTM that requires a CREB-mediated transcriptional response (Dubnau et al., 2003, Tully et al., 1994, Yin et al., 1995, Yin et al., 1994). It is also worth note that an alternative model has been proposed suggesting that these two forms of memory are mutually exclusive with ARM only forming after massed training and CREB dependent LTM only forming after spaced training (Isabel et al., 2004). In either case, we were interested to test whether consolidated memory after massed and spaced training differed in their sensitivity to our extinction protocol. We used these two training procedures to first condition wild-type animals and then 24 hours later, we subjected these flies to our extinction training procedure. As shown in Figure 2A, extinction significantly inhibited the memory performance of massed training group as well as the spaced training group (n=8, p<0.05). The effect of extinction after massed and spaced training were not significantly different because no extinction vs. training protocol interaction was observed (p>0.4). Thus consolidated ARM is clearly sensitive to extinction training. The fact that extinction had similar effects in both massed (ARM) and spaced groups argues that LTM also is sensitive to extinction as well. But the prevailing model in the literature suggests that memory after spaced training includes both ARM and LTM. Given this, the effectiveness of extinction training after each of these conditioning procedures also could be interpreted as an effect only on ARM. We tested this in two ways, one behavioral and the other genetic (see below).

Extinction of recent and remote memory

The interval between memory acquisition and the onset of extinction can have significant impacts on the long term effect of extinction in other behavior paradigms (Chang & Maren, 2009, Lopez et al., 2008, Maren & Chang, 2006, Myers et al., 2006, Rescorla, 2004, Schiller et al., 2008, Woods & Bouton, 2008). We tested this in our paradigm by comparing the immediate effects of extinction training applied 1-day or 4 days after spaced conditioning. In these experiments, animals were given 10 sessions of spaced training and then subjected to 30 cycles of extinction training either 1 day or 4 days later. In both cases, conditioned avoidance was tested immediately after the extinction training. Thus the 1-day group tested the effects of extinction on more recently acquired memory whereas the 4-day group tested the impact of extinction on remote memory. We observed a significant effect of extinction at both the 1-day and 4 day time-points. The effect of extinction training on 4 day (remote) memory retention was significantly less than with the 1 day (recent) memory retention time-point (Two-Way ANOVA, F(1, 88)=8.5, p<0.05) (Figure 2B). This is consistent with the hypothesis that short term effects of extinction training on remote memory is less severe than on recent memory, although this would need to be investigated in more detail.

The fact that we nevertheless observe significant extinction at the 4-day time-point also supports the hypothesis that CREB-dependent LTM per se can be extinguished (see above) because performance at this time-point is thought to be entirely composed of this form of memory (Tully et al., 1994, Yin et al., 1994).

Genetic investigation of extinction acquisition

Although a number of genes have been identified that underlie the consolidation of extinction (Quirk & Mueller, 2008), the molecular mechanisms underlying the acquisition of extinction are less understood. Our robust extinction paradigm in Drosophila, as well as a panel of genetic reagents that have been used to dissect acquisition and consolidation of Pavlovian conditioning, provided a unique opportunity to conduct a systematic genetic investigation of extinction. To begin such a dissection of extinction acquisition, we tested whether extinction learning and conditioned avoidance learning share common molecular mechanisms. To address this question, we tested effects on extinction of mutations in 5 different genes with established roles in Pavlovian olfactory conditioning. We selected from amongst a long list of single-gene mutations with known roles in Pavlovian memory based on 2 criteria: First, we identified mutations that reduced Pavlovian memory acquisition, but did not eliminate consolidated memory performance. This is important because effects of extinction on consolidated memory cannot practically be tested with mutants that eliminate consolidated memory in the Pavlovian task. Second, we selected mutations that span several signaling pathways with known roles in Pavlovian conditioning. The mutations that met these criteria were in rutabaga (a Ca++ responsive adenylyl cyclase), dNR1 (the Drosophila NMDA-R1 gene ), rugose (a PKA anchoring protein) and radish (a gene with novel sequence). We also tested a mutation in the amnesiac neuropeptide encoding gene, but residual levels of performance were so low even without extinction that the effects of extinction could not be reasonably assessed (data not shown). These mutations include two components of the cAMP signaling cascade (rutabaga and rugose), an NMDA-receptor, and a protein of unknown cellular function (radish).

rutabaga, which encodes a Ca++ responsive adenylyl cyclase, is widely thought to be a coincidence detector for the CS and US inputs (Abrams et al., 1998, Dubnau & Tully, 1998, Ferguson & Storm, 2004, Goodwin et al., 1997, Levin et al., 1992, Livingstone et al., 1984, Mons et al., 1999). A rutabaga null mutant, rutabaga1(rut1), exhibits approximately 50% of the normal memory performance both immediately after one training session (Tully & Quinn, 1985) and 24 hours after spaced training (Blum et al., 2009, Lu et al., 2007) (Figure 3A). We found that extinction learning appears to be normal in the rut1 mutants. With either 10 or 30 sessions of extinction, we observe significant inhibitory extinction in these mutants (Figure 3A, 3B). These findings are also consistent with those of Schwaerzel et al. (Schwaerzel et al., 2002) in which rutabaga mutants exhibited significant effects of extinction when it was applied shortly after a single Pavlovian training session.

rugose encodes a PKA anchoring protein and thus also is a component of the cAMP cascade (Shamloula et al., 2002). The rugose1 (rug1) mutant exhibits reduced STM but normal LTM (Zhong, personal communication). We find that as with rutabaga, extinction in rug1 appears to be normally manifested (Fig. 3C). We also tested extinction with mutations in amnesiac, which encodes a PACAP like neuropeptide and is thus postulated to be a component of the cAMP signaling cascade (Feany & Quinn, 1995). In this case, performance levels were so low after spaced training that we could not reasonably assess extinction (data not shown).

We next tested extinction with mutations in the Drosophila NMDA receptor 1, which is another potential coincidence detector for associative learning (Wu et al., 2007, Xia et al., 2005). We tested two independent dNR1 hypomorphic mutants, EP331 and EP3511 (Xia et al., 2005). Each of these mutant alleles exhibit reduced short-term and long-term memory. As with the rutabaga1 and rugose1 mutants, these dNR1 mutations exhibit normal levels of extinction learning (Fig. 3D, 3E, 3F).

Finally, we also tested extinction in animals mutant for the radish gene. radish encodes a novel gene and its normal function is required to support both short-term memory and the formation of ARM (Folkers et al., 1993, Folkers et al., 2006, Tully et al., 1994). Here too, we find that extinction learning with either 10 or 30 cycles of training was not different from wild type control (Fig. 3G, 3H).

rutabaga and NMDA-R1 dependent signaling account for the majority of Pavlovian learning performance

Each of the above mutants that we tested have robust defects in Pavlovian learning. On the other hand, each mutant also exhibits significant residual learning – which of course was essential for testing extinction effects on this residual memory. Thus none of these mutants on their own eliminate neural plasticity mechanisms underlying associative learning. Thus, while none of the mutants exhibit defects in extinction acquisition, we wondered if we accidently selected a group of genes that affect the same signaling pathway underlying Pavlovian learning. In this case Pavlovian and extinction learning might share a common parallel mechanism to that disrupted in these mutants. We addressed this possibility by measuring Pavlovian learning in a double mutant combination of rutabaga and dNR1. rutabaga and dNR1 are the two potential coincidence detectors that have known roles in this conditioning procedure, so we speculated that dNR1 function might account for some or all of the rutabaga independent conditioning. To test this hypothesis, we created rutabaga1, dNR1(EP331) double mutant animals and tested Pavlovian learning performance relative to that of each single mutant. As shown in Fig. 4, the double mutant combination exhibits significantly lower performance in our standard Pavlovian learning assay than either of the single mutants (n=18, p<0.05). This is consistent with the hypothesis that dNR1 represents a distinct mechanism of learning from rutabaga. It is worth note that although the rut1 mutant tested is believed to be a null allele (Levin et al., 1992, Livingstone et al., 1984), the EP331 allele of dNR1 is a likely hypomorph (Xia et al., 2005). Hence the low level of memory performance exhibited in the double mutant could either reflect residual dNR1 signaling, or a third mechanism of associative learning. Nevertheless, the extremely low level of performance in this rut1; dNR1 double mutant indicates that together, these two pathways account for the majority of the olfactory associative memory performance observed in this conditioning assay. The fact that extinction appears normal in each of these two mutants, as well as in the radish and rugose mutants, thus supports the hypothesis that extinction involves distinct cellular mechanisms.

Discussion

Extinction is a common feature of classical conditioning and appears to be conserved in Drosophila melanogaster (Lagasse et al., 2009, Schwaerzel et al., 2002, Tully & Quinn, 1985) as well as in other invertebrate species (Carew et al., 1981, Mccomb et al., 2002, Pedreira & Maldonado, 2003, Pedreira et al., 2004, Rankin, 2000, Richards et al., 1984, Stollhoff et al., 2005). In this study, we focused on the acquisition of extinction, which still is poorly understood at the cellular, biochemical and circuit levels. With the powerful genetic tools in Drosophila, memory can be genetically dissected, giving access to investigation of the underling molecular mechanisms and neural circuits. Many single-gene mutations have been identified that impact acquisition, consolidation or retrieval of memory (Heisenberg, 2003, Keene & Waddell, 2007, Margulies et al., 2005, Mcguire et al., 2005).

We designed a robust behavior paradigm to study extinction learning in Drosophila. Our approach is complementary to that of Mery and colleagues (Lagasse et al., 2009), but we used the standard electric shock reinforced Pavlovian paradigm instead of mechanical shock. This provided two advantages. First, this paradigm yields higher levels of performance, providing the possibility of observing partial effects. More importantly, most of the mutations affecting Pavlovian learning have been characterized only with the electro-shock version of this assay. As in Lagasse et al. (2009), extinction was performed one day after spaced or massed conditioning. However, memory performance then was tested immediately after odor exposure to examine the short term effects that are more likely to represent the acquisition of extinction. Our behavior paradigm also differs in several respects from earlier studies of extinction after conditioned reinforcement with electric-shock (Schwaerzel et al., 2002, Tully & Quinn, 1985). First, the onset of extinction training in this study is 24 hours after associative learning. In Drosophila aversive conditioning, the majority of short-term memory performance decays quickly during the first hours after training. Later on, memory consolidates into more stable forms that are resistant to disruption and to the passage of time. Thus we avoided potential interactions between memory decay and extinction consolidation. Studies in vertebrates also have shown that the interval between original memory acquisition with extinction onset can significantly impact the effect of extinction on memory retention (Chang & Maren, 2009, Lopez et al., 2008, Maren & Chang, 2006, Myers et al., 2006, Rescorla, 2004, Schiller et al., 2008, Woods & Bouton, 2008). By focusing on extinction of consolidated memory, we avoid this potential confound as well.

With this extinction procedure, we demonstrate significant extinction acquisition 24 hours after either spaced or massed training, suggesting that both forms of consolidated memory (ARM and LTM) are sensitive to this inhibitory extinction procedure. This notion is consistent with the observation that extinction also is effective when applied 4 days after memory acquisition.

Although the acquisition of extinction still is poorly understood, there are several reports where pharmacological or genetic perturbations impact extinction without disrupting classical conditioning (Cain et al., 2002, Marsicano et al., 2002). As an example, blockage of cannabinoid CB1 receptor by either gene knockout or CB1 antagonist both specifically impaired acquisition of extinction but left acquisition and consolidation of aversive memory intact (Marsicano et al., 2002). Consistently, cannabinoid reuptake inhibitors accelerate extinction learning (Chhatwal et al., 2005). On the other hand, there are counter examples where genetic machinery may be shared (Helmstetter & Fanselow, 1987, Mcnally & Westbrook, 2003, Sotres-Bayon et al., 2007). NMDA receptors (NMDArs), which are thought to be molecular coincidence detectors for associative learning, are essential for the acquisition of fear conditioning and other forms of conditioning (Barad & Cain, 2007). Studies of NMDArs blockade with a non-subunit selective antagonist, CPP ((±)-3-(2-carboxypiperazin-4-yl)-propyl-1-phosphonic acid), suggested that NMDA receptors plays a role in consolidation of extinction, rather than in acquisition (Santini et al., 2001, Suzuki et al., 2004). Ledoux and colleagues (Sotres-Bayon et al., 2007) showed on the other hand that ifenprodil, a NR2B subunit selective NMDArs antagonist of the NMDArs, blocked acquisition of extinction. This supports the view that NMDArs are required for acquisition of both associative conditioning and extinction.

We decided to address the relationships between mechanisms underlying acquisition of extinction versus Pavlovian conditioning. To do so, we examined extinction in a panel of Drosophila mutant strains with established roles in Pavlovian aversive learning. We chose mutants that span a range of signaling pathways. Among them are two members of the cAMP signaling cascade, an NMDA-receptor, and a protein of novel sequence. We believe that together, these genes represent signaling pathways that underlie the majority of performance in this Pavlovian task. rutabaga and NMDA receptor in particular are two genes that are thought to act as coincidence detectors in Drosophila associative learning (Heisenberg, 2003, Keene & Waddell, 2007, Margulies et al., 2005, Mcguire et al., 2005, Wu et al., 2007, Xia et al., 2005). rutabaga encodes a Ca⁺⁺ dependent adenylyl cyclase, and the rutabaga1 allele that we used is thought to be a null (Levin et al., 1992, Livingstone et al., 1984). The fact that rutabaga null alleles disrupt approximately 50% of learning performance suggests that there exists rutabaga independent learning pathway(s) (see also Blum et al., 2009). We hypothesized that these two coincidence detectors might have independent functions during short-term memory acquisition, and tested this by comparing the Pavlovian learning performance of rutabaga1; dNR1 (EP331) double mutants with that of each single mutant. Although double mutant animals still exhibit a small amount of residue performance (about 20% of that of wild type control), the Performance Index of this double mutant was significantly lower than that of each single mutant. This is consistent with the notion that dNR1 plays a role in rutabaga1 independent learning. Together, rutabaga, rugose, dNR1 and radish represent three different signaling pathways that underlie acquisition of Pavlovian aversive memory in flies, and as a group likely disrupt all or most of the signaling that can support “fear” conditioning. And yet each of these mutations leaves extinction intact. Together, our data are therefore consistent with the view that extinction and Pavlovian conditioning in Drosophila rely on distinct signaling pathways and/or neural circuits. We predict that mutations that disrupt extinction will likewise not affect acquisition of associative memory. This provides the foundation for efforts to identify extinction learning genes via genetic approaches in Drosophila.